Setup

First, we load packages, set the Google API key, and define a palette used for visualizing traffic data in leaflet.

## Load Google Traffic package
library(googletraffic)

## Load packages for working with and visualizing data
library(leaflet)
library(leaflet.extras)
library(leaflet.providers)
library(scales)
library(mapview)
library(raster)
library(tidyverse)

## Set Google API Key
google_key <- "GOOGLE-API-KEY-HERE"

## Leaflet Palette
pal <- colorNumeric(c("green", "orange", "red", "#660000"), 
                    1:4,
                    na.color = "transparent")

Key parameters

The following are key parameters used when querying Google Traffic data.

  • zoom: The zoom level defines the resolution of the traffic image. Values can range from 0 to 20. At the equator, with a zoom level 10, each pixel will be about 150 meters; with a zoom level 20, each pixel will be about 0.15 meters. Consequently, smaller zoom levels can be used if only larger roads are of interest (e.g., highways), while larger zoom levels will be needed for capturing smaller roads.
  • height/width: The height and width parameters define the height and width of the raster in terms of pixels. The kilometer height/width of pixels depends primarily on the zoom level (larger zoom levels correspond to the height and width having a smaller kilometer distance). Google traffic data takes time to render on a map, and larger height and width values require more time needed for data to render. The functions automatically scale the delay time depending on the height and width values set, but the delay time can also be manually set using the webshot_delay parameter. Note that traffic data may fail to render for very large height and width values, no matter the webshot_delay set.

Raster Around Point

The gt_make_raster() function produces a raster, using a centroid location and a height/width around the centroid to specify the location to query traffic information. The below example queries traffic for lower Manhattan, NYC.

## Make raster
r <- gt_make_raster(location   = c(40.712778, -74.006111),
                    height     = 1000,
                    width      = 1000,
                    zoom       = 16,
                    google_key = google_key)
#> Pausing for 5 seconds to allow traffic data to render

## Map raster
leaflet(width = "100%") %>%
  addProviderTiles("Esri.WorldGrayCanvas") %>%
  addRasterImage(r, colors = pal, opacity = 1)

By using a smaller zoom, we can capture a larger area; however, the pixels are more coarse.

## Make raster
r <- gt_make_raster(location   = c(41.384900, -78.891302),
                    height     = 1000,
                    width      = 1000,
                    zoom       = 7,
                    google_key = google_key)
#> Pausing for 5 seconds to allow traffic data to render

## Map raster
leaflet(width = "100%") %>%
  addProviderTiles("Esri.WorldGrayCanvas") %>%
  addRasterImage(r, colors = pal, opacity = 1)

Raster Around Polygon

The above example showed querying traffic information for lower Manhattan. Here, we show querying traffic information for all of Manhattan while still using a relatively high zoom level (that allows capturing traffic on smaller streets). The gt_make_raster_from_polygon() accepts a polygon as an input; if needed, multiple API queries are made to query traffic for the full polygon. We still specify the height and width, which determines the height and width used for a single API query. Large height and width values will result in fewer Google queries, while smaller height and width values will require more queries to cover the same spatial area; traffic data will fail to render if too large of height and width values are set.


## Grab polygon of Manhattan
us_sp <- getData('GADM', country='USA', level=2)
ny_sp <- us_sp[us_sp$NAME_2 %in% "New York",]

## Make raster
r <- gt_make_raster_from_polygon(polygon    = ny_sp,
                                 height     = 2000,
                                 width      = 2000,
                                 zoom       = 15,
                                 google_key = google_key)

## Map raster
leaflet(width = "100%") %>%
  addProviderTiles("Esri.WorldGrayCanvas") %>%
  addRasterImage(r, colors = pal, opacity = 1)

Raster Using Grid

gt_make_raster_from_polygon() creates a grid that covers a polygon, creates a traffic raster for each grid, and merges the rasters together. Some may prefer to first create and see the grid, then create a traffic raster using this grid. For example, one could (1) create a grid that covers a polygon then (2) remove certain grid tiles that cover areas that may not be of interest. The gt_make_grid() and gt_make_raster_from_grid() functions facilitate this process; gt_make_grid() creates a grid, then gt_make_raster_from_grid() uses a grid as an input to create a traffic raster.

First, we create a grid using gt_make_grid().

grid_df <- gt_make_grid(polygon = ny_sp,
                        height  = 2000,
                        width   = 2000,
                        zoom    = 15)

leaflet(width = "100%") %>%
  addTiles() %>%
  addPolygons(data = grid_df, popup = ~as.character(id))

We notice that the tile in the bottom left corner just covers water and some land outside of Manhattan. To reduce the number of API queries we need to make, we can remove this tile.

grid_clean_df <- grid_df[-5,]

leaflet(width = "100%") %>%
  addTiles() %>%
  addPolygons(data = grid_clean_df)

Second, we use the grid to make a traffic raster using gt_make_raster_from_grid().

## Make raster
r <- gt_make_raster_from_grid(grid_param_df = grid_clean_df,
                              google_key    = google_key)

## Map raster
leaflet(width = "100%") %>%
  addProviderTiles("Esri.WorldGrayCanvas") %>%
  addRasterImage(r, colors = pal, opacity = 1)

Make PNG then Convert to Raster

To make a google traffic raster, the functions first makes a temporary png file then converts the png file to a raster—where only the raster is outputted. Some workflows may require separating the processes. For example, if querying Google traffic data on a regular basis on a server, a user with a small server may want to minimize the processes done on the server. Here, a user could create and save PNG files on a server, then download the PNG files and convert the PNGs to a raster locally.

To support these workflows, the package provides the: * gt_make_png() function which creates a PNG file with traffic data * gt_load_png_as_traffic_raster() function which converts a PNG file into a spatially-referenced traffic raster

The below example illutrates the process.

#### Make png
# The function does not output anything in R; it saves a png file, specified using the "out_filename" parameter
gt_make_png(location = c(40.712778, -74.006111),
            height   = 1000,
            width    = 1000,
            zoom     = 16,
            out_filename = "google_traffic.png",
            google_key = google_key)

#### Convert png to raster
# We now convert the "google_traffic.png" created above into a raster. Because the png is not spatially referenced, we need to enter the same
r <- gt_load_png_as_traffic_raster(filename = "google_traffic.png",
                                   location = c(40.712778, -74.006111),
                                   height   = 1000,
                                   width    = 1000,
                                   zoom     = 16)

We can also use this process when querying traffic data for a larger study area that requires making multiple API calls. The below example illustrates creating multiple pngs from a grid.

#### Make grid
# We first make a grid, which contains all the parameters needed to make a png then the raster
grid_df <- gt_make_grid(polygon = ny_sp,
                        height  = 2000,
                        width   = 2000,
                        zoom    = 15)

print(grid_df)
#> Simple feature collection with 6 features and 6 fields
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -74.10988 ymin: 40.68462 xmax: -73.85324 ymax: 40.87875
#> Geodetic CRS:  WGS 84
#>   longitude latitude id height width zoom                       geometry
#> 1 -73.98156 40.84630  1   2000  2000   15 POLYGON ((-74.02448 40.8138...
#> 2 -73.89616 40.84630  2   2000  2000   15 POLYGON ((-73.93907 40.8138...
#> 3 -73.98156 40.78173  3   2000  2000   15 POLYGON ((-74.02448 40.7492...
#> 4 -73.89616 40.78173  4   2000  2000   15 POLYGON ((-73.93907 40.7492...
#> 5 -74.06696 40.71715  5   2000  2000   15 POLYGON ((-74.10988 40.6846...
#> 6 -73.98156 40.71715  6   2000  2000   15 POLYGON ((-74.02448 40.6846...
#### Make PNGs from grid
for(i in 1:nrow(grid_df)){
  grid_i_df <- grid_df[i,]
  
  gt_make_png(location     = c(grid_i_df$latitude, grid_i_df$longitude),
              height       = grid_i_df$height,
              width        = grid_i_df$width,
              zoom         = grid_i_df$zoom,
              out_filename = paste0(i, "_google_traffic.png"),
              google_key   = google_key)
}

#### Convert PNGs to rasters
# Here we make a list of rasters
r_list <- lapply(1 in 1:nrow(grid_df)){
  
  grid_i_df <- grid_df[i,]
  
  gt_load_png_as_traffic_raster(filename = paste0(i, "_google_traffic.png"),
                                location = c(grid_i_df$latitude, grid_i_df$longitude),
                                height   = grid_i_df$height,
                                width    = grid_i_df$width,
                                zoom     = grid_i_df$zoom)
}

#### Mosaic rasters together
# To mosaic the rasters together, the mosaic() function from the raster package requires that rasters have the same origin and resolution. The above rasters will not have the same orgin, and the resolutions will be slightly different. The gt_mosaic() function allows mosaicing rasters with different origins and resolutions.
r <- gt_mosaic(r_list)